Backup
NAS, SAN or Cloud Storage: Compare Cost, Access and Recovery
Local storage and cloud services solve different access and recovery problems. Compare complete costs, transfer-time maths and a practical hybrid design before choosing.

A team wants fast access to large project files in the office and easy collaboration outside it. Finance wants predictable spending. Management wants the business working again after a failure. These are connected requirements, but “local or cloud?” is too broad to answer until the workload, people and recovery target are clear.
NAS, SAN and Cloud Are Not Three Equivalent Boxes
NAS provides file access, commonly through SMB or NFS. SAN supplies block storage to servers through a supported fabric such as iSCSI or Fibre Channel. A server or application then manages the filesystem or data service. The SAN vs NAS guide explains that distinction in more detail.
Cloud storage describes a delivery model rather than one protocol. Cloud file services, block volumes, object storage and software-as-a-service platforms have different interfaces, performance characteristics, retention features and charges. An object-storage bucket is not automatically a drop-in replacement for an office SMB share.
| Option | Typical access model | Important dependency |
|---|---|---|
| Local NAS | Users and applications access shared files | NAS, local network, identity, power and backup |
| SAN | Hosts access block volumes | Supported hosts, storage paths, clustering and application design |
| Cloud file service | Managed network file access | Supported client path, identity, network and service limits |
| Cloud block storage | Volumes attached to supported cloud workloads | Cloud compute, application and recovery design |
| Cloud object storage | Objects accessed through applications or APIs | Application integration, request pattern and transfer policy |
Start with the Working Day
A video editor handling large local files, a branch user opening office documents and a cloud-hosted application do not need the same storage path. Map who creates the data, where it is used, how many people use it simultaneously and what happens when the internet is unavailable.
Local storage can keep suitable work inside the LAN, but local capacity alone does not provide reliable remote collaboration. A cloud service may make distributed access easier, but the client connection, latency and application behaviour still matter. Remote users need an intentional supported access method, not an exposed NAS administration port.
- List the active working set separately from older retained data.
- Identify large sequential transfers, many small files and latency-sensitive application requests.
- Record office, branch, home and mobile access requirements.
- Decide which work must continue during an internet or site outage.
Where Local Storage Can Be a Strong Fit
A suitably sized NAS can centralise permissions and keep large file traffic near the team. An all-flash model can suit a different access pattern from a capacity-focused HDD model. Check usable capacity, sustained disk performance, client connections and switch uplinks together.
The Synology DS925+ and ASUSTOR Flashstor 6 Gen 2 FS6806X illustrate different drive and connectivity designs. Their suitability depends on the application, supported drives, capacity and recovery plan; neither removes the need to assess the complete workload.
Use the NAS buying guide for bays and applications, and the storage speed and price guide for the data path. For SAN, include the supported host and path architecture in the cost and acceptance test.
Cloud Cost Is More Than Capacity per Month
Check the particular service, storage class and region. Costs may include stored data, operations or requests, retrieval, outbound transfer, minimum retention periods and the supporting compute or software. The applicable terms differ; not every service charges the same way.
A low-cost archive class may be appropriate for data rarely retrieved, while an active workflow needs a different response time and charging model. Backups, versions and replication may also hold more billed data than the current working folder suggests.
Model routine use and an unusually large recovery month. Include the cost of retrieving the data and the system needed to use it, not only the bill while it is sitting in storage.

Compare Five-Year Cost on the Same Basis
| Cost area | Local design | Cloud design |
|---|---|---|
| Initial setup | Hardware, drives, network changes and implementation | Migration, configuration, integration and identity setup |
| Monthly operations | Power, support, management and off-site protection | Storage, requests, transfer, software and management |
| Growth | Drives, expansion, additional nodes and migration work | Additional stored data, versions and service usage |
| Recovery | Replacement equipment, restore labour and independent copy | Retrieval, transfer, restore compute and recovery labour |
| Lifecycle | Refresh, support renewals and disposal | Service changes, exit transfer and migration |
A simple comparison is setup + 60 months of operating costs + growth and refresh + recovery and exit allowances. Use the same protected capacity, retention, service availability and support assumptions for both designs.
For the arithmetic alone, every US$100 of monthly recurring cost adds US$6,000 over 60 months. That is not a market price or a quotation. It shows why recurring software, support and backup charges belong beside the hardware cost in a five-year plan. Apply current taxes and currency assumptions consistently.
Restore-Time Maths: The Internet Can Be the Constraint
A recovery target needs more than a backup-completed message. If the data must cross an internet connection, available throughput sets a lower bound on transfer time.
| Data to transfer | Sustained transfer rate assumed | Ideal transfer time |
|---|---|---|
| 1 TB | 100 Mb/s | About 22 hours 13 minutes |
| 1 TB | 1 Gb/s | About 2 hours 13 minutes |
| 10 TB | 1 Gb/s | About 22 hours 13 minutes |
These calculations use decimal TB and eight bits per byte. They assume the stated rate is sustained and available to the transfer. Protocol overhead, competing traffic, retrieval preparation, small-file processing, decryption, application rebuild and validation add time. They are transfer calculations, not measured recovery results.
Define the recovery point objective, meaning how much recent work can be lost, and the recovery time objective, meaning how quickly a usable service must return. Then test the complete restore sequence against those targets.
A Practical Hybrid Pattern
A Hybrid Plan Needs a Tested Recovery Path
- Working Data
- Versioned Backup
- Off-site Copy
- Restore Test
Keep retention and deletion controls intentional. The working location and independent recovery copy can serve different needs.
One approach keeps the active working set on a suitable local system, creates versioned backups and retains an independent off-site copy. Another keeps the primary application in the cloud and backs it up to an appropriately separate target. The right direction depends on where the application runs and how the business recovers.
Separate access and deletion risks where the platform supports it. If one compromised administrator can delete the source, every snapshot and the only off-site copy, there is still a common failure path. Document retention, encryption-key custody and recovery-account access privately.
Synchronisation is useful for keeping working copies current. It may also propagate a deletion or damaged file. Check version retention and independent backup instead of counting two synchronised folders as two recovery plans.
Security and Data Location Need a Specific Design
Local storage is not automatically private or secure, and cloud storage is not automatically compliant. Both need appropriate identities, permissions, updates, logging, recovery controls and operational ownership. Confirm where the service holds data and backups, who can access them and which obligations apply to the organisation.
For remote work, use a supported secure access design and protect the endpoints that use the data. Avoid publishing credentials, recovery codes or private configuration details when asking for help. The backup restore-test guide provides a useful way to prove that a protected copy is actually usable.
Frequently Asked Questions
Is NAS cheaper than cloud storage?
It depends on capacity, use, retention, support, power, administration, backup and lifecycle costs. Compare the same service and recovery requirements over a defined period rather than hardware against a storage-only monthly rate.
Is cloud storage the same as NAS?
No. Cloud file services, block volumes, object storage and software platforms have different interfaces. Check how the application accesses data before treating one as a replacement for another.
Will a NAS work if the internet is down?
Suitable local services can continue over the LAN if power, local networking, identity and the application remain available. Remote users and internet-dependent functions may still be affected.
Does synchronising a NAS to the cloud count as backup?
It depends on the service and configuration. Synchronisation may propagate deletions or corruption. Check independent versions, retention, access separation and a tested restore process.
How long does it take to restore 1 TB from the cloud?
At a sustained 100 Mb/s, decimal 1 TB takes about 22 hours 13 minutes for transfer alone. At 1 Gb/s it takes about 2 hours 13 minutes. Retrieval, processing, overhead and validation add time.
Is local or cloud storage more secure?
Neither is automatically more secure. Identity, configuration, updates, logging, backup isolation, data location and operating practices determine the protection of the actual design.
Choose Access and Recovery Before Choosing Location
Choose local storage when its performance, control and operating model suit the workload; choose a cloud service when its delivery model fits the application and users. Use a supported SAN for the block-storage requirements that justify it. Compare hybrid options when the working copy and recovery copy have different needs.
Bring the active data size, growth, locations, applications and recovery targets to a Sales Specialist. A complete plan accounts for the working day, the monthly bill and the day something goes wrong.


